Transcript randomisation

Randomisation
James Erskine
Randomisation
You’re so random, compliment or insult?
 Randomisation is a mathematical /
statistical concept. It means there is no
detectable systematicity in an event
sequence. Thus, random refers to the lack of
structure or regularity. Sometimes there is
no regularity in behaviour hence - You’re so
random.

Types of Randomisation
Random sampling - this is distinct from
random allocation to experimental groups
but often confused. This may or may not
occur in experiments, most commonly it
does not.
 Random allocation to experimental
conditions. This is vital for a study to be
defined as a true experiment.
 Random ordering of experimental stimuli

Random? How?

Random sampling although nice rarely occurs.
Why? Because most of our samples are
convenience samples (the ubiquitous
undergraduates), where our samples are not of
an undergraduate flavour they tend to be clinical
groups we have specifically targeted (depressed,
anxious and so on). Neither are very random.
How do we become more
random?


Random allocation to groups is crucial.
Make sure there is no systematicity in the method
you allocate participants to groups by. Alternating
between your two experimental groups is clearly
not random. Putting 1 and 2 on bits of paper and
picking them out of a hat (yes it can be a bowl too)
is one method of becoming random. Or for the
more sophisticated we could use random number
tables in the back of most statistics books.
How do we become more
random continued 2

How do we use random number tables? Here we
just pick a number to start, lets say 7, then go to
line 7 and read the numbers from left to right, each
time you come across a 1 or 2 you allocate a
participant to that group, e.g. if the first two
numbers of line 7 are both 1 then participants 1
and 2 are both in group 1. Strictly speaking the
number 7 should also be chosen randomly.For a
web based random number generator go to:
www.random.org/nform.html
Randomness within
experimental tasks

Experimental stimuli should also be
random. E.g. if you are creating a lexical
decision task stimuli should be randomly
dispersed throughout the task. However this
will often not be possible depending on the
nature of your investigation. If a fixed order
must be used introduce as much
randomness as possible.
Conclusions
For experimenters being random is a great
compliment.
 We should all strive to be more random in
our experiments.

Useful Links
www.random.org/nform.html
 trochim.human.cornell.edu/kb/random.htm
